matplotlib-users

Re: [Matplotlib-users] Contouring a large set of ungridded data

[Jesper L. <jl...@dm...>]

On Saturday 15 October 2005 01:10, Jeff Whitaker wrote:
> Jesper:  Since this question has come up a couple of times, I decided to
> cook up an example.  First you'll need to download and install the
> natgrid python module (included in CDAT, but I've separated it out from
> the huge tarball and put it at
> ftp://ftp.cdc.noaa.gov/Public/jsw/natgrid.tar.gz).  Then try this:

Sounds great. I tried your example and it works fine (though I had to modify 
the import nat statement to "from natgrid import nat").

The only problem for my application is that I would like the interpolation to 
be limited to areas near observations and avoiding extrapolation to areas far 
from observations. Is that possible - from the natgrid documentation that 
does not seem to be the case?

If it is not possible, I will probably try to do it as a postprocessing of the 
interpolated data (probably simply check if the nearest observation is within 
a given range).

-- Jesper

Re: [Matplotlib-users] Contouring a large set of ungridded data

[Jeff W. <js...@fa...>]

Jesper Larsen wrote:
> On Saturday 15 October 2005 01:10, Jeff Whitaker wrote:
>   
>> Jesper:  Since this question has come up a couple of times, I decided to
>> cook up an example.  First you'll need to download and install the
>> natgrid python module (included in CDAT, but I've separated it out from
>> the huge tarball and put it at
>> ftp://ftp.cdc.noaa.gov/Public/jsw/natgrid.tar.gz).  Then try this:
>>     
>
> Sounds great. I tried your example and it works fine (though I had to modify 
> the import nat statement to "from natgrid import nat").
>
> The only problem for my application is that I would like the interpolation to 
> be limited to areas near observations and avoiding extrapolation to areas far 
> from observations. Is that possible - from the natgrid documentation that 
> does not seem to be the case?
>   

Jespers:  You can tell natgrid to not do any extrapolation outside the 
convex hull defined by the data points - see test.py in 
ftp://ftp.cdc.noaa.gov/Public/jsw/natgrid-0.1.tar.gz.  In that example I 
create a masked array that is only defined where there is no 
extrapolation needed.

-Jeff


-- 
Jeffrey S. Whitaker         Phone : (303)497-6313
NOAA/OAR/CDC  R/CDC1        FAX   : (303)497-6449
325 Broadway                Web   : http://www.cdc.noaa.gov/~jsw
Boulder, CO, USA 80305-3328 Office: Skaggs Research Cntr 1D-124

Re: [Matplotlib-users] Contouring a large set of ungridded data

[Eric F. <ef...@ha...>]

Chris and others,

Chris Barker wrote:
> Jeff Whitaker wrote:
> 
>> Andrea:  There are lots of things that would work well - but the 
>> licensing is a problem.  GTS is GPL, we need something with a less 
>> restrictive license (more like Python's).
> 
> 
> No, it's LGPL, which is much better, but maybe still not OK for mpl. I 
> guess the problem is that mpl is kind of a library of libraries, rather 
> than an application, so it would be awkward to have mixed licenses in it.
> 

I don't understand this.  Why can't the mpl license simply say that it 
applies to all components that do not cite other licenses, and then 
leave the reference to the original license in any code such as GTS 
which has another license?  This is not a plea for or against GTS or any 
other particular package, but rather an expression of puzzlement and 
frustration that we seem to be finding free software licenses limiting 
instead of liberating.

Eric

Re: [Matplotlib-users] Contouring a large set of ungridded data

[John H. <jdh...@ac...>]

>>>>> "Eric" == Eric Firing <ef...@ha...> writes:

    Eric> I don't understand this.  Why can't the mpl license simply
    Eric> say that it applies to all components that do not cite other
    Eric> licenses, and then leave the reference to the original
    Eric> license in any code such as GTS which has another license?
    Eric> This is not a plea for or against GTS or any other
    Eric> particular package, but rather an expression of puzzlement
    Eric> and frustration that we seem to be finding free software
    Eric> licenses limiting instead of liberating.

In my understanding, code which imports GPL'd code is obligated to
abide by the GPL restrictions itself if you want to distribute it,
specifically the copyleft part that says derived works must be open
source and distributed under the GPL.  So if mpl imports a GPL module,
mpl, then any code which is for distribution that imports matplotlib
is required to be distributed under the terms of the GPL.  This is
clearly unacceptable to someone who wants to distribute products that
use mpl under a proprietary license.  The LGPL was introduced to solve
this problem, and it basically says that you can import it and use it
but if you change the LGPL library the library changes must be
distributed under the LGPL.  I don't find it particularly
objectionable to use code which has an LGPL license, but given an
alternative, would rather have a BSD type license which is
"unencumbered".

Basically, the GPL is not particularly liberating, since it makes
strong statements about what you can do with code that uses GPL code.

JDH